Image Adaptive Watermarking

1
Image Adaptive Watermarking

• Ryan Johnson
• EE6886 Multimedia Security Systems
• Prof. Ching-Yung Lin
• May 10, 2006

2
Outline

• Project purpose and objectives
• Human vision system
• Image adaptive watermark scheme
• Experiments and results
• Conclusion

3
Project Purpose/Objectives

• Understand Human Vision System
• Human visual system
• Just Noticeable Difference (JND)
• Compare IA watermarking schemes
• DCT vs. Spread Spectrum Technique (Cox et al)
• Do proposed IA watermarking schemes and JND
  calculations agree to real world
  simulation/evaluation?

4
Human Vision System

• Broad ranging/complex topics
• Eye and brain interaction Nava Rubin, New York
  University
• DCT/Wavelet quantization Andrew Watson, NASA
• Frequency sensitivity
• Luminance sensitivity
• Contrast masking

5
Human Vision System

• Frequency Sensitivity
• Eyes sensitivity to sine wave gratings at
  different frequencies
• Watson argues that there is a specific sine
  wave grating that matches the human contrast
  detector.
• Image independent
• single matrix generated independent of any image
• Based on the smallest coefficient that would
  yield a visible signal

6
Human Vision System

• Luminance Sensitivity
• Measure of detectable noise on a constant
  background
• Can adapt to 1010 different intensity levels,
  but not simultaneously
• Adjusts overall sensitivity in order to do this
  called brightness adaptation
• Non-linear function
• Image dependent
• Uses frequency sensitivity and knowledge of the
  mean luminance of the display

7
Human Vision System

• Contrast Masking
• Delectability of one signal in the presence of
  another signal
• Interrelated to luminance sensitivity and
  dependent on luminance
• Image dependent
• Uses knowledge of the luminance sensitivity

8
Just Noticeable Difference

• Webers Law
• JND threshold

9
Image Adaptive Watermarking

• Podilchuk et. al. from Bell Laboratories
• Proposed two techniques
• Discrete Cosine Transform
• Wavelet Transform
• Improvement over spread spectrum technique
• Improvement in images with large uniform areas

10
Image Adaptive Watermarking

• Frequency sensitivity
• Standard JPEG quantization table

11
Image Adaptive Watermarking

• Luminance Sensitivity

12
Image Adaptive Watermarking

• Contrast Masking
• Result is a JND matrix based on the input image

13
Image Adaptive Watermarking

• Only apply watermarking coefficients where image
  DCT coefficients are larger than JND thresholds

14
Experiments and Results
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
Experiments and Results

• Capacity
• Spread Spectrum 1000 watermark coefficients

21
Experiments and Results

• Transparency
• PSNR
• MSSIM measurement
• Structural Similarity
• A more direct way to compare structures
  watermarked and unwatermarked images
• Errors are perceived changes in structural
  information variation
• Mimics the HVS

22
Experiments and Results

• Robustness
• Resize and rescale attack sub sampling and
  interpolation
• JPEG compression compression
• Averaging filter lowpass filtering
• Unsharp filter edge enhancement

23
Experiments and Results

• JND Test
• Multiplied the JND matrix by different percentages

24
Conclusion

• Human Vision System
• Just noticeable differences
• Image Adaptive Watermarking
• Maximizes watermark using knowledge of HVS
• Other things to check out
• Wavelet Transform IA Watermarking
• Watermark detection without original image
• Tamper detection
• Different types of watermarking sequences
• Etc

25
Questions?
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)